Triggering a saccade toward a static visual target depends upon spatial attributes. The smaller the target eccentricity (or saccade amplitude), the longer the latency. According to the equilibrium hypothesis, a goal directed eye movement is not initiated as long as the visuo-oculomotor system is within a mode where opposing commands counterbalance each other. We further investigated this dependence in macaques trained to track a target which, after moving toward the central visual field, moved away in the same direction. During such conditions, the target eccentricity diminishes during the centripetal motion and increases during the centrifugal portion. After fixating a central static target for a variable interval, a peripheral target appeared at 8 possible locations and moved toward the symmetrical location in the opposite visual field. The monkey was rewarded for tracking it until its disappearance. For each trial, the target path was pseudo-randomly selected and its speed was non constant: the target decelerated from a maximum speed (40, 60 or 80°/s) to rest. The different target decelerations were tested during different blocks of trials. After target onset, the eye often drifted in the same direction as the target motion. The glissade amplitude increased with time. For the slowest target, the glissade started before the target crossed the center of the display. By contrast, it started after the central crossing when the target moved faster. Saccade latency increased while the centripetal target approached gaze direction. As the target became centrifugal, the saccade latency increased much more. For the fastest target, centrifugal saccades were predominant and exhibited an opposite tendency insofar as the small target eccentricities (small saccades) were no longer associated with the longest latencies. More generally, our study confirms that saccade triggering does not come down to a temporal process which is free from spatial attributes of gaze or target.